Neural network construction via a linear program

Author

Pollatschek, M.A.

Author_Institution

Fac. of Ind. Eng. & Manage., Technion-Israel Inst. of Technol., Haifa, Israel

fYear

1995

fDate

7-8 March 1995

Abstract

A procedure with polynomial complexity is proposed to construct a feed-forward neural network with binary inputs and a single binary output. The problem is formulated as one in linear inequalities and is solved by a modification of the primal simplex algorithm. We claim that it finds an acceptably low number of artificial neurons in the hidden layer or, equivalently, the number of separating hyperplanes.

Keywords

backpropagation; feedforward neural nets; linear programming; multilayer perceptrons; artificial neurons; binary inputs; feedforward neural network; hidden layer; linear inequalities; linear program; neural network construction; polynomial complexity; primal simplex algorithm; separating hyperplanes; single binary output; Artificial neural networks; Biological neural networks; Engineering management; Feedforward neural networks; Feedforward systems; Industrial engineering; Neural networks; Neurons; Polynomials; Technology management;

fLanguage

English

Publisher

ieee

Conference_Titel

Electrical and Electronics Engineers in Israel, 1995., Eighteenth Convention of

Conference_Location

Tel Aviv, Israel

Print_ISBN

0-7803-2498-6

Type

conf

DOI

10.1109/EEIS.1995.514163

Filename

514163

Link To Document

https://search.isc.ac/dl/search/defaultta.aspx?DTC=49&DC=2350316